Re: [patch v3 31/36] x86/apic: Provide cpu_primary_thread mask

[Thomas Gleixner <tglx@xxxxxxxxxxxxx>]

On Wed, May 24 2023 at 23:48, Kirill A. Shutemov wrote:
> On Mon, May 08, 2023 at 09:44:17PM +0200, Thomas Gleixner wrote:
>>  #ifdef CONFIG_SMP
>> -/**
>> - * apic_id_is_primary_thread - Check whether APIC ID belongs to a primary 
>> thread
>> - * @apicid: APIC ID to check
>> - */
>> -bool apic_id_is_primary_thread(unsigned int apicid)
>> +static void cpu_mark_primary_thread(unsigned int cpu, unsigned int apicid)
>>  {
>> -    u32 mask;
>> -
>> -    if (smp_num_siblings == 1)
>> -            return true;
>>      /* Isolate the SMT bit(s) in the APICID and check for 0 */
>> -    mask = (1U << (fls(smp_num_siblings) - 1)) - 1;
>> -    return !(apicid & mask);
>> +    u32 mask = (1U << (fls(smp_num_siblings) - 1)) - 1;
>> +
>> +    if (smp_num_siblings == 1 || !(apicid & mask))
>> +            cpumask_set_cpu(cpu, &__cpu_primary_thread_mask);
>>  }
>> +#else
>> +static inline void cpu_mark_primary_thread(unsigned int cpu, unsigned int 
>> apicid) { }
>>  #endif
>>  
>>  /*
>
> This patch causes boot regression on TDX guest. The guest crashes on SMP
> bring up.

I rather call it a security feature: It makes TDX unbreakably secure.

> The change makes use of smp_num_siblings earlier than before: the mask get
> constructed in acpi_boot_init() codepath. Later on smp_num_siblings gets
> updated in detect_ht().
>
> In my setup with 16 vCPUs, smp_num_siblings is 16 before detect_ht() and
> set to 1 in detect_ht().

  early_init_intel(c)
    if (detect_extended_topology_early(c) < 0)
       detect_ht_early(c);

  acpi_boot_init()
    ....

  identify_boot_cpu(c)
    detect_ht(c);

Aaargh. That whole CPU identification code is a complete horrorshow.

I'll have a look....